Call for Assistance - udev

[Alf mel]

On Thursday 23 August 2007 09:48:25 pm Alexander E. Patrakov wrote:
> I would still like udev to be optional.

I agree.

> Also, udev and bootscripts are
> closely tied to each other (things like mounting /proc, /sys, /dev,
> triggering uevents, remounting root read-write, copying generated rules,
> retriggering failed uevents should happen in the right order). Thus, it
> would be just insane to add udev and not bootscripts (or at least some
> words about them). I will help if we decide what to do with the
> bootscripts and how to deal with non-SysV inits.

Perhaps it would be useful to come up with a  "boot sequence description" 
that would list what things should be done and in what order.  It wouldn't 
be actual scripts; just something that would talk about what a good boot 
sequence should do.

One advantage from udev is the ability to probe for hardware.  If the kernel 
is built with modules and with automatic module loading support, udev can 
probe for the hardware, load necessary modules, populate /dev and configure 
the devices.

Another thing to remember is that udev does not have to use any kind of ram 
disk to work.  That means device nodes could be permanently written to disk 
once and left alone.  For servers this would be particularly useful since 
they hardly change their  hardware configuration.

> It is enough to
> create the following devices, directories and symlinks for a full LiveCD
> build: null, console, zero, ptmx, tty, random, urandom, fd, stdin,
> stdout, stderr, pts, shm, core. Then the reader can decide at the end,
> whether he wants to use udev or static /dev.

You can also use /proc/partitions to build the device nodes for your disks 
and partitions.  That's what my bootscripts do before they mount 
partitions.  Once partitions are mounted, I run udev and have it find the 
remaining hardware and do its thing.

-- 
@ - Alf